Secure-by-Design AI: Protecting MLOps in the Microsoft Cloud with Martin Dimovski [MVP-MCT]

Beskrivelse

In this episode of the m365.fm podcast, Mirko Peters sits down with Microsoft MVP, MCT, cloud security expert, and community leader Martin Dimovski to explore one of the most important topics in modern enterprise IT: securing AI workloads and MLOps environments inside the Microsoft Cloud. Together, they dive deep into secure-by-design architecture, AI security risks, DevSecOps, Prompt Injection attacks, identity protection, Microsoft Defender, GitHub Advanced Security, and the future of AI-driven cyber threats. Martin shares his personal journey from IT support engineer into cloud security and AI security architecture, explaining how years of experience in infrastructure, Azure, DevOps, and Microsoft technologies ultimately pushed him toward cybersecurity and AI governance. The discussion highlights why AI security is no longer optional and why organizations that move too fast without proper security foundations could face major problems in the coming years. WHY AI SECURITY MATTERS NOW MORE THAN EVER One of the strongest themes throughout this episode is the speed at which organizations are deploying AI systems without fully understanding the security implications behind them. Martin explains that many companies are currently: * Deploying AI solutions rapidly * Experimenting with LLM integrations * Building AI agents * Creating cloud-native AI workloads * Using open-source AI models * Integrating APIs into production environments But at the same time, organizations often forget the security fundamentals that should protect these environments. The conversation explores how AI introduces completely new attack surfaces while simultaneously amplifying existing security problems. WHAT “SECURE-BY-DESIGN” REALLY MEANS A major focus of the episode is understanding the concept of secure-by-design architecture. Martin explains that security should never be added after development is complete. Instead, security conversations must begin at the very first design phase of any application or AI project. The discussion covers: * Threat modeling * Architectural reviews * Identity security * Authentication planning * Secure pipelines * Infrastructure protection * Secure APIs * Data governance Martin shares why collaboration between developers, architects, DevOps engineers, and security teams is absolutely essential for building resilient AI systems. One of the key takeaways: Security teams should not become blockers for innovation — they should become partners in building secure systems. UNDERSTANDING MLOPS & DEVSECOPS For listeners newer to AI infrastructure topics, Martin breaks down the differences between: * DevOps * DevSecOps * MLOps * Secure AI pipelines The episode explains how machine learning operations combine infrastructure, automation, data engineering, model deployment, and monitoring into one continuous operational process. Martin also highlights why traditional security approaches are no longer enough once organizations start integrating: * Large Language Models * AI agents * Cloud AI services * AI APIs * AI orchestration pipelines The discussion shows how modern security must now cover not only infrastructure and applications, but also models, prompts, training data, inference pipelines, and AI-generated outputs. THE REAL DANGER OF PROMPT INJECTION One of the most fascinating parts of the episode is Martin’s explanation of Prompt Injection attacks. Using simple real-world analogies, Martin explains how attackers manipulate Large Language Models by overriding or bypassing original system instructions. The conversation explores: * Direct Prompt Injection * Indirect Prompt Injection * AI manipulation * LLM instruction abuse * Malicious prompts * Unsafe AI agents * Context hijacking * Data extraction risks Martin explains why prompt injection is becoming one of the most discussed attack vectors in AI security today and why organizations need to start thinking about AI trust boundaries immediately. THE HIDDEN RISK OF OPEN-SOURCE MODELS Another major topic is the increasing use of publicly available AI models. Martin shares concerns around: * Downloading unverified models * Compromised Hugging Face repositories * Malicious AI packages * Unsafe dependencies * Supply-chain attacks * API key exposure * Secret leakage * Public model poisoning The discussion highlights how organizations may unknowingly introduce compromised models directly into production environments. This section serves as a major warning for companies rushing into AI adoption without proper governance and validation processes. WHY IDENTITY SECURITY IS EVERYTHING Identity and access management become another core theme throughout the episode. Martin strongly emphasizes the importance of: * Microsoft Entra ID * Privileged Identity Management * Just-In-Time access * Least privilege * Identity governance * Access reviews * Role separation * Conditional Access One of the strongest lessons from the conversation is that attackers often do not need to break systems — they simply abuse existing permissions and weak access configurations. Martin explains why organizations should avoid giving permanent privileged access and instead embrace short-lived administrative permissions wherever possible. MICROSOFT DEFENDER & AI SECURITY The episode also dives deeply into the Microsoft security ecosystem and how Microsoft Defender is evolving to protect AI workloads. Martin discusses: * Microsoft Defender for Cloud * Defender XDR * AI workload monitoring * Real-time scanning * Azure AI Foundry protection * Threat visibility * Security telemetry * Cloud-native protection According to Martin, Microsoft Defender is becoming one of the most powerful unified security platforms for organizations heavily invested in Microsoft technologies. Become a supporter of this podcast: https://www.spreaker.com/podcast/m365-fm-modern-work-security-and-productivity-with-microsoft-365--6704921/support [https://www.spreaker.com/podcast/m365-fm-modern-work-security-and-productivity-with-microsoft-365--6704921/support?utm_source=rss&utm_medium=rss&utm_campaign=rss].

Stop Building Bots, Start Building Runtimes: A Field Guide to Microsoft Agents

Everyone is calling Build 2026 the AI conference. Most of the attention went toward new copilots, voice experiences, and increasingly capable models. But beneath the headlines, Microsoft quietly introduced something far more significant. The real story is not about another AI feature. It is about the emergence of a completely new infrastructure layer for enterprise computing. For years, organizations approached AI as a chatbot problem. Build a conversational interface, connect it to some data, add a few prompts, and call it an AI strategy. That approach worked for experimentation, but it was never designed for scale. Chatbots forget context, struggle with governance, and become increasingly difficult to manage as more departments begin building their own solutions. What Microsoft is building now is fundamentally different. We are moving from assistants that answer questions to agents that operate as active participants inside the enterprise. THE FOUR-LAYER MODEL THAT CHANGES EVERYTHING One of the most important concepts emerging from Microsoft's latest announcements is the idea that agents should no longer be viewed as products. They should be viewed as layers within a larger system. Most organizations currently evaluate AI by comparing products. They ask whether they should use Copilot, Copilot Studio, Azure AI Foundry, GitHub Copilot, or Security Copilot. That approach creates confusion because these technologies solve very different problems. The better way to think about agents is through architecture. The modern agent stack consists of four distinct layers: * Experience Layer * Agent Layer * Runtime Layer * Governance Layer Each layer serves a unique purpose. Each layer has different stakeholders. And each layer introduces different operational requirements. Organizations that understand this distinction can scale successfully. Organizations that ignore it often end up with fragmented deployments and duplicated effort. WHY IDENTITY IS THE REAL STORY The most important announcement from Build 2026 was not a new agent. It was identity. Historically, automation systems operated through shared service accounts. Scripts, bots, and integrations all ran under generic credentials that nobody really owned. This created security blind spots and made auditing nearly impossible. When something happened, it was difficult to determine which system actually performed the action. Microsoft's new model changes that entirely. Every agent now receives its own identity inside Microsoft Entra. Every agent becomes a first-class principal within the organization. It has its own permissions, its own audit trail, and its own lifecycle. This seemingly small architectural change creates enormous downstream benefits: * Least-privilege access * Full auditability * Conditional Access enforcement * Individual credential management * Instant revocation capabilities For the first time, agents are being treated like actual actors inside the enterprise rather than invisible background processes. This shift enables governance at a scale that simply wasn't possible before. THE RISE OF AGENT INFRASTRUCTURE Most organizations are still focused on building individual agents. The problem is that individual agents are only part of the story. Real business value emerges when agents work together. A retrieval agent gathers information. An analysis agent interprets it. A communication agent creates output. A coordinating agent manages the workflow. Suddenly, what looked like a chatbot becomes an operational system. This is where Azure AI Foundry Agent Service enters the picture. Foundry provides the runtime environment where agents actually execute. It handles: * Memory management * Session persistence * Multi-agent orchestration * Tool discovery * State management Instead of developers spending months building infrastructure, they can focus on defining agent behavior while Microsoft manages scaling, networking, and execution behind the scenes. This dramatically reduces complexity and accelerates deployment timelines. THE SHADOW AGENT PROBLEM One of the most fascinating challenges discussed in this episode is something many organizations have not yet recognized. The Shadow Agent problem. Building agents is becoming incredibly easy. Governance is not. As a result, business units increasingly create their own agents without involving IT. Sales teams build lead qualification agents. Operations teams create workflow automations. Individual departments experiment with Copilot Studio and Power Platform. Before long, dozens or even hundreds of agents are operating across the organization without centralized visibility. This creates significant risks: * Duplicate functionality * Excessive permissions * Compliance concerns * Data leakage risks * Lack of ownership Agent 365 is Microsoft's answer to this challenge. It provides centralized discovery, governance, identity management, auditing, and policy enforcement across the entire agent ecosystem. The goal is not to stop innovation. The goal is to make innovation manageable. FROM ASSISTANCE TO AUTOMATION The biggest change is not technical. It is organizational. For years, AI systems were designed to assist humans. The human remained the primary actor while AI provided recommendations and suggestions. The new generation of agents flips that relationship. The agent executes. The human supervises. Sales qualification becomes automated. Security triage becomes automated. Financial reconciliation becomes automated. Humans focus on judgment, strategy, relationships, and decision-making while agents handle repetitive operational work. This fundamentally changes how organizations think about productivity. Instead of helping employees complete tasks faster, agents begin completing entire categories of tasks on their own. Humans shift toward oversight, governance, and exception handling. THE FUTURE ISN'T MORE CHATBOTS Build 2026 may ultimately be remembered as the moment agents stopped being experimental technology and started becoming enterprise infrastructure. The organizations that succeed over the next decade will not be the ones with the most chatbots. They will be the ones that understand identity, governance, orchestration, runtime architecture, and multi-agent systems. They will build platforms rather than isolated tools. The future of enterprise AI is not conversational. The future of enterprise AI is operational. And Microsoft has just laid the foundation for that future. Become a supporter of this podcast: https://www.spreaker.com/podcast/m365-fm-modern-work-security-and-productivity-with-microsoft-365--6704921/support [https://www.spreaker.com/podcast/m365-fm-modern-work-security-and-productivity-with-microsoft-365--6704921/support?utm_source=rss&utm_medium=rss&utm_campaign=rss].

2. juli 20261 h 16 min

Microsoft Copilot Adoption: What Actually Works - With Chris Hinch [Microsoft]

Artificial Intelligence has moved beyond experimentation and into the heart of modern business. Yet while organizations are investing heavily in Microsoft Copilot, many struggle to achieve meaningful adoption and measurable business value. Simply assigning licenses is no longer enough. Successful AI transformation requires governance, training, executive sponsorship, security, and a well-defined adoption strategy that helps employees integrate AI into their daily work. In this episode, Microsoft Cloud Solution Architect Chris Hinch shares practical lessons learned from working with enterprise customers adopting Microsoft Copilot at scale. Together, we separate marketing hype from real-world implementation and explore what organizations should focus on to maximize productivity, improve employee satisfaction, and build a sustainable AI culture. WHY MOST COPILOT DEPLOYMENTS STRUGGLE Many organizations approach Microsoft Copilot expecting immediate productivity gains. They purchase licenses, enable the service, and assume employees will naturally discover how to use AI effectively. Unfortunately, this approach often leads to disappointing adoption rates and limited return on investment. Chris explains that AI is not a magic solution capable of fixing broken business processes overnight. Like any enterprise technology, Copilot requires clear objectives, structured onboarding, continuous learning, and organizational leadership. Companies that define measurable business outcomes before deployment consistently achieve stronger adoption than those implementing AI simply because it is the latest technology trend. ADOPTION IS A PEOPLE CHALLENGE, NOT A TECHNOLOGY CHALLENGE Technology rarely becomes the biggest obstacle during deployment. Instead, successful adoption depends on helping employees change how they work. Every department has unique workflows, challenges, and productivity goals, making a one-size-fits-all rollout ineffective. Rather than deploying Copilot across the entire organization immediately, Chris recommends identifying practical business problems that AI can solve quickly. Demonstrating measurable improvements builds confidence, encourages wider adoption, and creates internal momentum for future AI initiatives. Successful adoption strategies include: * Department-specific use cases * Clear business objectives * Continuous employee training * Executive sponsorship * Ongoing success measurement THE POWER OF CHAMPIONS PROGRAMS One of the most effective strategies discussed in this episode is establishing an internal Champions Program. Instead of relying solely on IT departments, organizations identify enthusiastic employees from different business units who become early adopters and advocates for Microsoft Copilot. These champions experiment with prompts, discover practical workflows, and share successful techniques with colleagues. Their real-world experience makes AI more approachable than traditional technical documentation or generic training sessions. As adoption grows, these internal experts naturally become trusted advisors who accelerate organizational learning while reducing resistance to change. PROMPTING IS ABOUT CONTEXT, NOT COMPLEXITY The conversation also explores one of the biggest misconceptions surrounding AI—prompt engineering. Rather than memorizing complicated prompt structures, users should focus on providing meaningful context. Chris explains Microsoft's simple prompting framework, emphasizing goals, context, available information, and expected outcomes. AI produces significantly better responses when users explain why they need something instead of simply asking for a task to be completed. Whether summarizing emails, creating presentations, analyzing documents, or generating reports, context consistently improves the quality and relevance of AI-generated responses. COPILOT, COPILOT STUDIO, AND AI FOUNDARY Microsoft's AI ecosystem continues expanding rapidly, which often creates confusion about the different products available. This episode breaks down where Microsoft Copilot, Copilot Studio, Agent Builder, and Azure AI Foundry fit within an enterprise AI strategy. Organizations beginning their AI journey should focus on end-user productivity with Microsoft Copilot before gradually expanding into custom agents and enterprise automation through Copilot Studio. As maturity increases, Azure AI Foundry enables more advanced AI scenarios involving custom models, orchestration, and enterprise-grade AI development. Core AI technologies discussed include: * Microsoft Copilot * Copilot Studio * Agent Builder * Azure AI Foundry * Microsoft 365 Copilot Chat SECURITY, GOVERNANCE, AND TRUST Security remains one of the most common concerns organizations raise before deploying AI. Chris explains that Microsoft Copilot respects existing Microsoft 365 permissions, meaning users can only access information they already have permission to view. At the same time, AI frequently exposes governance weaknesses that already exist within organizations. Poor SharePoint permissions, excessive file sharing, outdated ownership, and inconsistent access controls become much more visible when AI begins searching organizational content. Rather than creating new security risks, Copilot often highlights governance issues that should have been addressed long before AI entered the organization. MICROSOFT PURVIEW, ENTRA ID, AND DEFENDER Enterprise AI adoption extends well beyond productivity tools. Microsoft Purview, Microsoft Entra ID, Microsoft Defender, and SharePoint Advanced Management all play essential roles in creating secure AI environments. These technologies allow organizations to classify sensitive information, enforce access policies, monitor AI usage, detect Shadow AI, prevent unauthorized data sharing, and ensure compliance across Microsoft 365. Important governance capabilities include: * Data classification * Identity management * Shadow AI detection * Information protection * Secure AI governance THE FUTURE OF MICROSOFT COPILOT Looking ahead, Chris shares his excitement about Microsoft's rapid AI innovation, including Copilot enhancements, advanced PowerPoint generation, collaborative AI experiences, Agent capabilities, Microsoft Scout, and expanding Model Context Protocol (MCP) support. Rather than replacing employees, future Copilot experiences will increasingly automate repetitive work, orchestrate complex business processes, generate sophisticated business assets, and assist knowledge workers throughout their daily workflows. As AI becomes more deeply integrated into Windows, Microsoft 365, and enterprise applications, organizations that invest today in governance, training, and adoption strategies will be best positioned to capitalize on these emerging capabilities. FINAL THOUGHTS Microsoft Copilot adoption is not simply an IT deployment—it is an organizational transformation that combines technology, leadership, governance, security, and continuous learning. As Chris Hinch explains throughout this conversation, organizations achieve the greatest success when they focus first on solving real business problems rather than deploying AI for its own sake. With strong executive sponsorship, Champions Programs, practical training, secure governance, and department-specific use cases, Microsoft Copilot becomes far more than another productivity tool. It becomes a trusted digital assistant that helps employees reclaim time, improve collaboration, reduce repetitive work, and unlock the full potential of AI across the modern workplace. Become a supporter of this podcast: https://www.spreaker.com/podcast/m365-fm-modern-work-security-and-productivity-with-microsoft-365--6704921/support [https://www.spreaker.com/podcast/m365-fm-modern-work-security-and-productivity-with-microsoft-365--6704921/support?utm_source=rss&utm_medium=rss&utm_campaign=rss].

30. juni 202654 min

Secure-by-Design AI: Protecting MLOps in the Microsoft Cloud with Martin Dimovski [MVP-MCT]

Beskrivelse

Kommentarer

Prøv gratis i 14 dager

Alle episoder